Re: [Xen-devel] [PATCH v3 2/9] xen-block: add document for mutli hardware queues/rings

["Roger Pau Monné" <roger.pau@citrix.com>]

El 02/10/15 a les 18.12, Wei Liu ha escrit:
> On Fri, Oct 02, 2015 at 06:04:35PM +0200, Roger Pau Monné wrote:
>> El 05/09/15 a les 14.39, Bob Liu ha escrit:
>>> Document multi queues/rings of xen-block.
>>>
>>> Signed-off-by: Bob Liu <bob.liu@oracle.com>
>>
>> As said by Konrad, you should send this against the Xen public headers
>> also (or even before). I have a comment below.
>>
>>> ---
>>>  include/xen/interface/io/blkif.h |   32 ++++++++++++++++++++++++++++++++
>>>  1 file changed, 32 insertions(+)
>>>
>>> diff --git a/include/xen/interface/io/blkif.h b/include/xen/interface/io/blkif.h
>>> index c33e1c4..b453b70 100644
>>> --- a/include/xen/interface/io/blkif.h
>>> +++ b/include/xen/interface/io/blkif.h
>>> @@ -28,6 +28,38 @@ typedef uint16_t blkif_vdev_t;
>>>  typedef uint64_t blkif_sector_t;
>>>  
>>>  /*
>>> + * Multiple hardware queues/rings:
>>> + * If supported, the backend will write the key "multi-queue-max-queues" to
>>> + * the directory for that vbd, and set its value to the maximum supported
>>> + * number of queues.
>>> + * Frontends that are aware of this feature and wish to use it can write the
>>> + * key "multi-queue-num-queues", set to the number they wish to use, which
>>> + * must be greater than zero, and no more than the value reported by the backend
>>> + * in "multi-queue-max-queues".
>>> + *
>>> + * For frontends requesting just one queue, the usual event-channel and
>>> + * ring-ref keys are written as before, simplifying the backend processing
>>> + * to avoid distinguishing between a frontend that doesn't understand the
>>> + * multi-queue feature, and one that does, but requested only one queue.
>>> + *
>>> + * Frontends requesting two or more queues must not write the toplevel
>>> + * event-channeland ring-ref keys, instead writing those keys under sub-keys
>>> + * having the name "queue-N" where N is the integer ID of the queue/ring for
>>> + * which those keys belong. Queues are indexed from zero.
>>> + * For example, a frontend with two queues must write the following set of
>>> + * queue-related keys:
>>> + *
>>> + * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
>>> + * /local/domain/1/device/vbd/0/queue-0 = ""
>>> + * /local/domain/1/device/vbd/0/queue-0/ring-ref = "<ring-ref>"
>>> + * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn>"
>>> + * /local/domain/1/device/vbd/0/queue-1 = ""
>>> + * /local/domain/1/device/vbd/0/queue-1/ring-ref = "<ring-ref>"
>>> + * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn>"
>>
>> AFAICT, it's impossible by design to use multiple queues together with
>> multipage rings, is that right?
>>
> 
> As far as I can tell, these two features are not inherently coupled.
> Whether you want to make (by design) them coupled together or not is
> another matter. :-)

I haven't looked at the implementation yet, but some mention of whether
multipage-rings are allowed with multiqueue would be good. For example
if both can indeed be used in conjunction I would mention:

If multi-page rings are also used, the format of the grant references
will be:

/local/domain/1/device/vbd/0/queue-0/ring-ref0 = "<ring-ref0>"
/local/domain/1/device/vbd/0/queue-0/ring-ref1 = "<ring-ref1>"
/local/domain/1/device/vbd/0/queue-0/ring-ref2 = "<ring-ref2>"
[...]

Roger.